Relative Clause Extraposition and Prosody

References

Relative Clause Extraposition and Prosody

Claudia Poschmann and Michael Wagner

Goethe-Universitat Frankfurt a.M. / McGill University, Montreal

6th of March, 2014

Claudia Poschmann and Michael Wagner Relative Clause Extraposition and Prosody

References

Extraposition of (Restrictive) Relative Clauses

(1) a. Peter hat jemanden besucht, der krank ist. (RRC)‘Peter has visited someone who is ill.’

b. Peter hat niemandem gesagt, dass er krank ist. (CC)‘Peter didn’t tell anybody that he is ill.’

(2) a. Peter hat jemanden, der krank ist, besucht. (RRC)Peter has visited someone who is ill.’

b. *Peter hat niemandem, dass er krank ist, gesagt. (CC)‘Peter didn’t tell anybody that he is ill’


References

(Some) Factors affecting Extraposition of (R)RCs:

I Length (of the RRC):Longer RCs tend to be extraposed. (e.g Cullicover andJackendoff, 2005)

I Distance (between RRC and Head)The acceptability of RCE is inversely proportional to thedistance between RC and head. (e.g. Hemforth et al., 2000;Uszkoreit et al., 1998)

I Distance and Length interact:If distance is increased, even longer RCs tend to stay in situ.(e.g. Hemforth et al., 2000; Uszkoreit et al., 1998)

I What is distance?Number of intervening words / syllables / new d-refs (...)?


References

(Some) Factors affecting Extraposition of (R)RCs:

Discourse Focus:

I Rochemont and Culicover (1990), Takami (1999):Extraposition tends to occur when an RRC is in focus andexpresses new information, while the matrix-VP is discoursegiven.

I Shannon (1992): Extraposition is more likely if the head ofthe RC is focused than if it represents the discourse topic.

I If the head is focused, subsequent material is backgrounded.


References

Prominence of the intervening material

Hypothesis I: (Contextual Prominence)RC-extraposition improves if the intervening material is part of thebackground.

Hypothesis II: (Prosodic Prominence)Extraposability correlates inversely with the prosodic prominence ofintervening material.

Problem: How can we tease apart Hypothesis I and II?


References

Extraposition and RC-Type

(3) (Emonds, 1979, p.234)

a. Some men appeared at the door that Mary had beeninsulting. (RRC)

b. *These men appeared at the door, who Mary had beeninsulting. (ARC)

c. These men, who Mary had been insulting, appeared atthe door. (ARC)

I Strong Adjacency Requirement for ARCsHigh Syntactic Attachment (Emonds, 1979; McCawley,1981): ARCs have to co-indexed with the head at the surface.Bi-dimensional Logic (Potts, 2005a): Appositive Contentcannot be moved.

I Consequence: Most of the previous studies only investigatedthe extraposition of RRCs.


References

But: Discourse Relations matter

(4) (Holler, 2005, p.150)

a. IhreTheir

Lehrerinteacher

wolltenwanted

diethe

Kinderchildren

besuchen,visit,

diewho

aberPART

nichtnot

zuat

Hausehome

war.was.

‘The children wanted to visit their teacher, who wasnot at home.’

b. IhreTheir

Lehrerin,teacher,

diewho

aberPART

nichtnot

zuat

Hausehome

war,was,

wolltenwanted

diethe

Kinderchildren

besuchen.visit.

‘The children wanted to visit their teacher, who wasnot at home.’


References

Moreover: Distance, Length and Focus matter

(5) (Arnold, 2007, p.288)

a. Someone came who Mary knew. [RRC]b. ?John came, who Mary knew. [ARC]c. Even John came, who everyone had expected would be

too scared of potential publicity.

ARC Extraposition improves if ...

I ... distance is kept minimal. (Holler, 2005)

I ... the ARC is made heavier.(Arnold, 2007)

I ... the head of the ARC is focused. (Heringa 2012)


References

RC-Type and Extraposition

Hypothesis III:Strong Version: ARCs do not extrapose at all.Weak Version: ARCs are harder to extrapose than RRCs.


References

RC-Type and Prosody

I ARCs are prosodically less integrated than RRCs.

I ARCs have a strong boundary intonation(comma-intonation).(Selkirk, 2004; Potts, 2005b)

I RRCs form part of the focus-background-structure of thematrix clause.

I ARCs have their own focus- background-structure. (Holler,2005; Riester, 2009)


References

RC-Type and Prosody

I No Focus-Projection from ARC to matrix-clause

(6) Which sister did Peter call?

a. Peter called MARIA, who is living in HAMBURG.b. ?Peter called Maria, who is living in HAMBURG.c. Peter called the sister who is living in HAMBURG.

I No Association with Focus between matrix-clause and ARC

(7) a. Peter only called Maria, who is CARLA’s bestfriend.

b. Peter called Maria, who is only CARLA’s bestfriend.


References

Interaction of RC-Type and Focus

Hypothesis IV: Focus and RC-TypeThe effects of Focus and RC-Type on WordOrder interact.


References

Experiments

Design:

I Number of Participants: 35

I Number of Experiments: 2

I Number of Items: 18

I Number of Conditions: 6

Factors:

I RC-Type (ARC / RRC)

I Focus (Object / Subject / Wide)

I WordOrder (extraposed / non-extraposed)

Type of Task:

I Production-Experiment

I Acceptability-Test (scale 1 - 7)


References

Example for a Testitem with RRC

(8) a. War die Wanderung schwierig?‘Was the hike difficult?’ (Wide-Focus)

b. Wer hat das Riemannhaus erreicht?‘Who reached the Riemann house?’ (Subject-Focus)

c. Welches Ziel haben die Wanderer erreicht?‘Which goal did the hiker reach?’ (Object-Focus)

(9) a. (Nein,) jeder Wanderer, der Schneeschuhe trug, hatdas Riemannhaus erreicht.‘(No,) every hiker who was wearing snow shoes hasreached the Riemannhaus.’

b. (Nein,) jeder Wanderer hat das Riemannhaus erreicht,der Schneeschuhe trug.‘(No,) every hiker has reached the Riemannhaus, whowas wearing snow shoes.’


References

Example for a Testitem with ARC

(10) a. War die Wanderung schwierig?‘Was the hike difficult?’ (Wide-Focus)

b. Wer hat das Riemannhaus erreicht?‘Who reached the Riemann house?’ (Subject-Focus)

c. Welches Ziel hat der Wanderer erreicht?‘Which goal did the hiker reach?’ (Object-Focus)

(11) a. (Nein,) der Wanderer, der ja Schneeschuhe trug,hat das Riemannhaus erreicht.‘(No,) the hiker, who was wearing snow shoes, hasreached the Riemannhaus.’

b. (Nein,) der Wanderer hat das Riemannhaus erreicht,der ja Schneeschuhe trug.‘(No,) the hiker has reached the Riemannhaus, whowas wearing snow shoes.’


References

Expected Focus-Pattern

(12) Subject-Focus:A: Wer hat das Riemannhaus erreicht?‘Who reached the Riemann house?’B: Der WANDERER hat das Riemannhaus erreicht, der jaSchneeschuhe trug.‘The HIKER has reached the Riemannhaus, who waswearing snow shoes.’

(13) Object-Focus:A: Welches Ziel hat der Wanderer erreicht?‘Which goal did the hiker reach?’B: Der Wanderer hat das RIEMANNHAUS erreicht, der jaSchneeschuhe trug.‘The hiker has reached the RIEMANNHAUS, who waswearing snow shoes.’


References

Step 1: Acceptability-Test

Predictions:

I Hypothesis I (Contextual Prominence):Subject-Focus > Wide Focus > Object-Focus

I Hypothesis II (Prosodic Prominence):Subject-Focus > Wide Focus > Object-Focus

I Hypothesis III (RC-Type):extraposed RRCs > extraposed ARCs

I Hypothesis IV (Interaction of RC-Type and Focus):RC-Type and Focus interact


References

Results Acceptability-Test

2

4

6

8

RRC ARCType

Acc

epta

bilit

y R

atin

g

●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●

2

4

6

8

Extraposed Non−ExtraposedWordOrder

Acc

epta

bilit

y R

atin

g

2

4

6

8

Wide Object SubjectFocus

Acc

epta

bilit

y R

atin

g

Figure : Responses by WordOrder, Focus, and RC-Type.


References


Wide Object Subject

●●●●

●●●●

●●●● ●●●●●●●●

2

4

6

8

2

4

6

8

RR

CA

RC

Extraposed Non−Extraposed Extraposed Non−Extraposed Extraposed Non−ExtraposedWordOrder

resp

onse WordOrder

ExtraposedNon−Extraposed

Figure : Responses by WordOrder, Focus, and RC-Type.


References


Wide Object Subject

●●●● ●●●● ●●●● ●●●●●●●●

2

4

6

8

2

4

6

8

Extraposed

Non−

Extraposed

RRC ARC RRC ARC RRC ARCType

resp

onse WordOrder

ExtraposedNon−Extraposed

Figure : Responses by RC-Type, Focus, and Wordorder


References


Table : Extraposability by RC-type, Focus, and WordOrder

Dependent variable:Naturalness Rating

TypeRRC.vs.ARC 0.067 (0.055)WordOrderExtraposed.vs.Non-Extraposed −0.475∗∗∗(0.069)FocusSubject.vs.Other 0.057 (0.060)FocusWide.vs.Object 0.108 (0.069)RRC.vs.ARC:Extraposed.vs.Non-Extraposed 0.045 (0.039)RRC.vs.ARC:FocusSubject.vs.Other 0.042 (0.041)RRC.vs.ARC:FocusWide.vs.Object −0.027 (0.047)Extraposed.vs.Non-Extraposed:FocusSubject.vs.Other 0.272∗∗∗ (0.041)Extraposed.vs.Non-Extraposed:FocusWide.vs.Object 0.027 (0.047)RRC.vs.ARC: Extraposed.vs.Non-Extraposed:FocusSubject.vs.Other −0.127 (0.081)RRC.vs.ARC: Extraposed.vs.Non-Extraposed:FocusWide.vs.Object −0.014 (0.095)Constant −0.013 (0.043)

Observations 1,127

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References


Table : Results for Extraposed RCs

Dependent variable:Naturalness Rating

RCRestrictive.vs.Non-Restrictive 0.087 (0.060)FocusSubject.vs.Other 0.197∗∗∗ (0.049)FocusWide.vs.Object 0.138∗∗∗ (0.051)RRC.vs.ARC:FocusSubject.vs.Other −0.038 (0.088)RRC.vs.ARC:FocusWide.vs.Object −0.060 (0.143)Constant −0.250∗∗∗ (0.051)Observations 552

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References


Table : Results for Non-Extraposed RCs

Dependent variable:

Naturalness Rating

TypeRRC.vs.ARC 0.012 (0.063)FocusObject −0.104∗∗∗ (0.035)FocusSubject −0.128∗∗ (0.054)TypeRRC.vs.ARC:FocusObject 0.015 (0.084)TypeRRC.vs.ARC:FocusSubject 0.111 (0.108)Constant 0.299∗∗∗ (0.047)

Observations 575

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References

Main-Findings Acceptability-Test

I Significant Effect of WordOrderIn all conditions, extraposed RCs rated lower thannon-extraposed RCs

I Significant Interaction of Focus and WordOrderUnder Extraposition, Subject-Focus rated higher than Wideand Wide Focus rated higher than Object-Focus

I No Interaction of RC-Type and WordOrderExtraposed ARCs rated as high as extraposed RRCs

I No Interaction of RC-Type and FocusBut with in situ ARCs, Subject-Focus rated lower than Objectand Wide Focus.


References

Step 2: Evaluation of the Acoustic Data

I Hypothesis II:Can we tease apart the effects of Focus and ProsodicProminence?

I Hypothesis IV:Can we find any interaction between RC-Type and Focus?


References

RC Extraposition and Prosodic Prominence

Extraposed Non−Extraposed

0%

25%

50%

75%

100%

0%

25%

50%

75%

100%

RR

CA

RC

Wide Object Subject Wide Object SubjectFocus

coun

t VPAccentVP unaccentedVP accented

Figure : Proportion of utterances with unaccented VPs.


References

RC Extraposition and Prosodic Prominence

Table : Logistic Regression Model Testing for Effects on ProsodicProminence

Dependent variable:

VP Accentuation

WordOrderExtraposed.vs.Non-Extraposed −2.477∗∗∗ (0.418)Subject.vs.Other −4.309∗∗∗ (0.443)Wide.vs.Object 0.198 (0.583)RRC.vs.ARC −0.499 (0.389)Extraposed.vs.Non-Extraposed: Subject.vs.Other −0.390 (0.633)Extraposed.vs.Non-Extraposed: Wide.vs.Object −0.390 (1.111)Extraposed.vs.Non-Extraposed:RRC.vs.ARC −1.443∗ (0.760)FocusSubject.vs.Other: RRC.vs.ARC 0.938 (0.618)FocusWide.vs.Object: RRC.vs.ARC 0.827 (1.105)Extraposed.vs.Non-Extraposed:Subject.vs.Other:RRC.vs.ARC 1.960 (1.231)Extraposed.vs.Non-Extraposed:Wide.vs.Object:RRC.vs.ARC 0.343 (2.210)Constant 2.596∗∗∗ (0.289)

Observations 1,133

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References

RC-Extraposition and Prosodic Prominence

Extraposed Non−Extraposed

●●

2

4

6

2

4

6

RR

CA

RC

VP unaccented VP accented VP unaccented VP accentedVPAccent

resp

onse

WordOrderExtraposedNon−Extraposed

Figure : Acceptability rating in subject focus, both in extraposed andnon-extraposed word orders.


References


Table : The Effect of Prominence on Naturalness in Subject Focus

Dependent variable:

Naturalness Rating

VPunaccented.vs.VPaccented 0.001 (0.063)RRC.vs.ARC 0.137 (0.099)Extraposed.vs.Non-Extraposed −0.320∗∗∗ (0.084)VPunaccented.vs.VPaccented:RRC.vs.ARC −0.093 (0.092)VPunaccented.vs.VPaccented: Extraposed.vs.Non-Extraposed 0.305∗∗∗ (0.098)RRC.vs.ARC: Extraposed.vs.Non-Extraposed 0.011 (0.081)VPunaccented.vs.VPaccented: RRC.vs.ARC:Extraposed.vs.Non-Extraposed −0.297∗ (0.173)Constant −0.045 (0.055)

Observations 378

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References


RRC ARC

●

●

●

●

●

●

●●

●

●

●

●

●

●

●

●●

●●●

●

●

●

●●

●

●

●

●●●

●

●●

●●●

●

●●●

●

●●●

●

●●

●

●

●●●

●

●

100

200

300

100

200

300

Extraposed

Non−

Extraposed

Wide ObjectSubject Wide ObjectSubject

Max

imum

Pitc

h on

the

Obj

ect

RRC ARC

●

●

60

70

80

90

60

70

80

90

Extraposed

Non−

Extraposed

Wide Object Subject Wide Object Subject

Inte

nsity

of t

he O

bjec

t

Figure : Maximum pitch (Hz) and maximum intensity (dB) on the object.


References


Table : Effect of Prosodic Prominence and Focus on Naturalness

Dependent variable:

Naturalness

cObjectPitch −0.205 (0.125)cObjectIntensity −0.147∗∗ (0.065)FocusSubject.vs.Other 0.0001 (0.095)FocusWide.vs.Object 0.089 (0.183)Constant 0.025 (0.053)

Observations 1,047

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References


I Hypothesis II: The Acceptability of RC- Extraposition isinversely proportional to the Prosodic Prominence of theintervening material.

I Hypothesis IV: In non-extraposed word-order, we find asignificant interaction between RC-Type and Focus.


References

Comma-Intonation

Wide Object Subject

●

●

●

●●

●●

●

●

●●

●

0.25

0.50

0.75

1.00

0.25

0.50

0.75

1.00

Extraposed

Non−

Extraposed

RRC ARC RRC ARC RRC ARC

Leng

th o

f Wor

d P

rece

ding

RC

Figure : Duration of the word preceding the RC in non-extraposed andextraposed word order.


References

Comma-Intonation

Table : The length of the Word Preceding the RC.

Dependent variable:

z-score of log duration

Extraposed.vs.Non-Extraposed −0.285∗∗ (0.144)TypeRRC.vs.ARC −0.123∗∗∗ (0.036)FocusSubject.vs.Other 0.004 (0.022)FocusWide.vs.Object 0.001 (0.019)Extraposed.vs.Non-Extraposed:RRC.vs.ARC 0.159∗∗∗ (0.024)Extraposed.vs.Non-Extraposed:FocusSubject.vs.Other −0.099∗∗∗ (0.025)Extraposed.vs.Non-Extraposed:FocusWide.vs.Object −0.017 (0.029)RRC.vs.ARC:FocusSubject.vs.Other −0.037 (0.025)RRC.vs.ARC:FocusWide.vs.Object −0.030 (0.029)Extraposed.vs.Non-Extraposed:RRC.vs.ARC:FocusSubject.vs.Other 0.041 (0.050)Extraposed.vs.Non-Extraposed:RRC.vs.ARC:FocusWide.vs.Object −0.009 (0.058)Constant 0.019 (0.125)

Observations 1,047

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References

Fall or Rise?

Wide Object Subject

●●

●

●

●

●●

●

●

●

100

150

200

250

300

350

100

150

200

250

300

350

Extraposed

Non−

Extraposed

RRC ARC RRC ARC RRC ARC

Pitc

h at

Bou

ndar

y

Figure : Mean pitch of the last quadrant of the word preceding the RC.


References

Fall or Rise?

Table : Mean pitch of the last quadrant of the word preceding the RC

Dependent variable:

z-score of mean pitch

WordOrderExtraposed.vs.Non-Extraposed −0.236∗∗∗ (0.072)TypeRRC.vs.ARC 0.053∗∗ (0.024)FocusSubject.vs.Other −0.049 (0.030)FocusWide.vs.Object 0.019 (0.057)WordOrderExtraposed.vs.Non-Extraposed:TypeRRC.vs.ARC 0.087∗∗∗ (0.031)WordOrderExtraposed.vs.Non-Extraposed:FocusSubject.vs.Other 0.103∗∗∗ (0.032)WordOrderExtraposed.vs.Non-Extraposed:FocusWide.vs.Object 0.018 (0.038)TypeRRC.vs.ARC:FocusSubject.vs.Other 0.112∗∗∗ (0.032)TypeRRC.vs.ARC:FocusWide.vs.Object −0.019 (0.038)WordOrderExtraposed.vs.Non-Extraposed:TypeRRC.vs.ARC:FocusSubject.vs.Other −0.100 (0.065)WordOrderExtraposed.vs.Non-Extraposed:TypeRRC.vs.ARC:FocusWide.vs.Object −0.005 (0.075)Constant 0.003 (0.071)

Observations 995

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01


References

Conclusions:

I Acceptability of RC extraposition is inversely proportional tothe prominence of the material intervening between head andRC.

I Acceptability-ratings correlate not just with the contextualsalience, but also with prosodic prominence.

I ARCs and RRCs are equally natural when extraposed. Thischallenges approaches which assume a strict adjacencyrequirement for ARCs (e.g. Potts 2005).

I Under extraposition, no interaction between RC-Type andFocus.

I In the non-extraposed case, however, significant interactionswere found in the prosodic data, which shows that naturalnessof an ARC decreases significantly if it separates accented fromunaccented material.


References

THANK YOU!

This work was supported by a grant of the DFG Research Group”Relativsatze”, Frankfurt.


References

Doug Arnold. Non-restrictive relatives are not orphans. Journal ofLinguistics, 43:271–309, 2007.

Peter Cullicover and Ray Jackendoff. Simpler Syntax. OxfordUniversity Press, Oxford, 2005.

Joseph Emonds. Appositive relatives have no properties. LinguisticInquiry, 10:211–243, 1979.

Barbara Hemforth, Lars Konieczny, and Christoph Scheepers.Modifier attachment: relative clauses and coordinations. InBarbara Hemforth and Lars Konieczny, editors, German sentenceprocessing, pages 159–163. Kluwer, Dordrecht, 2000.

Anke Holler. Weiterfuhrende Relativsatze. Empirische undtheoretische Aspekte. Akademie Verlag, Berlin, 2005.

James D. McCawley. The syntax and semantics of english relativeclauses. Lingua, 53:99–149, 1981.

Chris Potts. Lexicalized intonational meaning. In ShigetoKawahara, editor, Papers on Prosody, volume 30 of University ofMassachussetts Occasional Papers in Linguistics, pages 129–146.GLSA, Amherst, Ma., 2005a.


References

Chris Potts. The Logic of Conventional Implicatures. OxfordUniversity Press, Oxford, 2005b.

Arndt Riester. Stress test for relative clauses. In Arndt Riester andEdgar Onea, editors, Focus at the Syntax-Semantics Interface,volume 3 of Working Papers of the SFB 732. University ofStuttgart, 2009.

Michael Rochemont and Peter Culicover. English focusconstructions and the theory of grammar. Cambridge UniversityPress, Cambridge, 1990.

Elisabeth Selkirk. Comments on intonational phrasing in english.In Marina Vigario Sonia Frota and Maria Joao Freitas, editors,Prosodies. Mouton de Gruyter, 2004.

Thomas Shannon. Toward an adequate characterization of relativeclause extraposition in modern german. In Irmengard Rauch,Gerald F. Carr, and Robert L. Kyes, editors, On GermanicLinguistics. Issues and Methods, pages 253–281. Mouton deGruyter, Berlin/New York, 1992.


References

Ken-ichi Takami. A functional constraint on extraposition from np.In Akio Kamio and Ken ichi Takami, editors, Function andStructure, pages 23–56. John Benjamins, Amsterdam, 1999.

Hans Uszkoreit, Thorsten Brants, and al. Studien zurperformanzorientierten linguistik: Aspekte derrelativsatzextraposition im deutschen. CLAUS Report No. 991–14, Universitat des Saarlandes, Saarbrucken, 1998.


Date post:	10-Jan-2022
Category:	Documents
Upload:	others
View:	2 times
Download:	0 times

Relative Clause Extraposition and Prosody

Documents